In [ ]:
"""
 We were using the MTSD dataset : https://www.kaggle.com/datasets/zeuss2k3/mapillary-traffic-sign-dataset
 but its too heavy .
 -----------
 So we decided to use : https://www.kaggle.com/datasets/nomihsa965/traffic-signs-dataset-mapillary-and-dfg
"""

NOTE : For running this notebook you should download the dataset from :¶

https://www.kaggle.com/datasets/nomihsa965/traffic-signs-dataset-mapillary-and-dfg¶

and then put it in the main directory (Argos/) by "data" naming¶

In [ ]:
# Input data files are available in the read-only "../data/" directory

# import os
# for dirname, _, filenames in os.walk('/kaggle/input'):
#     for filename in filenames:
#         print(os.path.join(dirname, filename))

Explore Dataset Argos Project: Traffic Sign Detection¶

University of Tehran

Author: MahdiHaji

Subject: CPS

In [3]:
import numpy as np
import pandas as pd
import os
import cv2
import matplotlib.pyplot as plt
import glob

import shutil
import sys
from IPython.display import display
from joblib import Parallel, delayed
from tqdm.notebook import tqdm 
tqdm.pandas()
from sklearn.model_selection import train_test_split

📖 Meta Data¶

  • crops/ - Folder containing images of traffic signs that have been cut from the original images and divided into 76 folders corresponding to 76 types of traffic signs.

  • images/ - Folder containing set photos of the form image_name.jpg

  • txts (YOLO)/ - All label for photos of the form image_name.txt, each label is form

[label object] [x_center] [y_center] [width] [height] with scale (0,1)

  • xmls/ - All label for photos of the form image_name.txt

  • classes.json - containing a list of all classes with information about class name, class index and number of instances in the dataset.

  • dataset.yaml - yolo's dataset settings

✏️ Write Images¶

  • We need to copy the Images to Current Directory(/kaggle/working) as /kaggle/input doesn't have write access which is needed for YOLOv5.
  • We can make this process faster using Joblib which uses Parallel computing.
In [21]:
## @title Get image name
PATH = '../data'
list_img=[img for img in os.listdir(PATH+'/images')]
list_txt=[img for img in os.listdir(PATH+'/txts (YOLO)')]

path_img=[]
for i in range (len(list_img)):
    path_img.append(PATH+'/images/'+list_img[i])


path_label=[]
for i in range (len(list_txt)):
    arr = [PATH+'/txts (YOLO)/'+list_txt[i], ]
    path_label.append(PATH+'/txts (YOLO)/'+list_txt[i])
In [22]:
df = pd.DataFrame(path_img)
In [23]:
df
Out[23]:
0
0 ../data/images/0010010.jpg
1 ../data/images/-1ZXiNNunp0TA-puToNbeQ.jpg
2 ../data/images/TtH4hYk8DSfmgb0omrdUWA.jpg
3 ../data/images/0000203.jpg
4 ../data/images/Jw6diCiDRA_s7QJ71hy-Yg.jpg
... ...
19341 ../data/images/7KpbGD6xXhocK83WLKU6aA.jpg
19342 ../data/images/0010009.jpg
19343 ../data/images/JXvJr1vwTqwvKH7vBMeVrg.jpg
19344 ../data/images/0007575.jpg
19345 ../data/images/XHRRyeCq2sGi6gQi1cFmtA.jpg

19346 rows × 1 columns

In [24]:
dfl = pd.DataFrame(path_label)
In [25]:
dfl
Out[25]:
0
0 ../data/txts (YOLO)/0005047.txt
1 ../data/txts (YOLO)/0015254.txt
2 ../data/txts (YOLO)/0003436.txt
3 ../data/txts (YOLO)/0004359.txt
4 ../data/txts (YOLO)/0013625.txt
... ...
19043 ../data/txts (YOLO)/0000126.txt
19044 ../data/txts (YOLO)/0014153.txt
19045 ../data/txts (YOLO)/3qf88IGrJRd8pNl1ShzqQQ.txt
19046 ../data/txts (YOLO)/0012522.txt
19047 ../data/txts (YOLO)/0002731.txt

19048 rows × 1 columns

Make DataFrame image and label¶

In [54]:
PATH_TXT = "../data/txts (YOLO)/"
metadata = {'image': [],
            'x_center': [],
            'y_center': [],
            'w': [],
            'h': [],
            'class': [],
            'num_labels': []}

for file in path_label:
    f = open(file, "r")
    labels = f.read().split('\n')[:-1]
    filerp = file[:]
    for label in labels:
    
        lis = label.split(" ")
        metadata['num_labels'].append(len(labels))
        metadata['image'].append(PATH+"/images/"+file.split("/")[-1][:-3]+'jpg')
        metadata['x_center'].append(float(lis[1]))
        metadata['y_center'].append(float(lis[2]))
        metadata['w'].append(float(lis[3]))
        metadata['h'].append(float(lis[4]))
        metadata['class'].append(int(lis[0]))
In [55]:
df = pd.DataFrame(metadata)
In [56]:
df
Out[56]:
image x_center y_center w h class num_labels
0 ../data/images/0005047.jpg 0.790365 0.401852 0.078646 0.142593 28 2
1 ../data/images/0005047.jpg 0.795573 0.248148 0.077604 0.138889 21 2
2 ../data/images/0015254.jpg 0.887592 0.524272 0.031941 0.061489 59 2
3 ../data/images/0015254.jpg 0.809275 0.476537 0.048526 0.064725 1 2
4 ../data/images/0003436.jpg 0.788281 0.508796 0.169271 0.056481 47 1
... ... ... ... ... ... ... ...
30553 ../data/images/3qf88IGrJRd8pNl1ShzqQQ.jpg 0.707465 0.515873 0.005704 0.009921 0 1
30554 ../data/images/0012522.jpg 0.462224 0.287217 0.123464 0.165049 50 3
30555 ../data/images/0012522.jpg 0.234029 0.254854 0.128993 0.139159 46 3
30556 ../data/images/0012522.jpg 0.833538 0.095874 0.100737 0.125405 18 3
30557 ../data/images/0002731.jpg 0.765365 0.303241 0.136979 0.236111 58 1

30558 rows × 7 columns

x_center vs y_center¶

In [57]:
from scipy.stats import gaussian_kde


x_val = df['x_center'].values
y_val = df['y_center'].values

# Calculate the point density
xy = np.vstack([x_val,y_val])
z = gaussian_kde(xy)(xy)

fig, ax = plt.subplots(figsize = (10, 10))
# ax.axis('off')
ax.scatter(x_val, y_val, c=z, cmap='viridis')
# ax.set_xlabel('x_mid')
# ax.set_ylabel('y_mid')
plt.show()
No description has been provided for this image

width Vs height¶

In [58]:
w_val = df['w'].values
h_val = df['h'].values

# Calculate the point density
wh = np.vstack([w_val,h_val])
z = gaussian_kde(wh)(wh)

fig, ax = plt.subplots(figsize = (10, 10))
# ax.axis('off')
ax.scatter(w_val, h_val, c=z, cmap='viridis')
# ax.set_xlabel('bbox_width')
# ax.set_ylabel('bbox_height')
plt.show()
No description has been provided for this image

Number of each class¶

In [59]:
cls = df['class'].value_counts()
x = np.arange(len(cls))
y = []
for i in range(len(cls)):
    y.append(cls[i])
plt.bar(x,y)
Out[59]:
<BarContainer object of 76 artists>
No description has been provided for this image

🌈 Visualization¶

In [67]:
PATH_TXT = "../data/txts (YOLO)"
metadata = {'image': [],
            'label': []}

for i, file in enumerate(path_label):
    f = open(file, "r")
    labels = f.read().split('\n')[:-1]
    filel = file.split('/')
    metadata['image'].append(PATH+'/images/'+filel[3][:-3]+'jpg')
    metadata['label'].append(file)
In [68]:
df = pd.DataFrame(metadata)
In [69]:
df
Out[69]:
image label
0 ../data/images/0005047.jpg ../data/txts (YOLO)/0005047.txt
1 ../data/images/0015254.jpg ../data/txts (YOLO)/0015254.txt
2 ../data/images/0003436.jpg ../data/txts (YOLO)/0003436.txt
3 ../data/images/0004359.jpg ../data/txts (YOLO)/0004359.txt
4 ../data/images/0013625.jpg ../data/txts (YOLO)/0013625.txt
... ... ...
19043 ../data/images/0000126.jpg ../data/txts (YOLO)/0000126.txt
19044 ../data/images/0014153.jpg ../data/txts (YOLO)/0014153.txt
19045 ../data/images/3qf88IGrJRd8pNl1ShzqQQ.jpg ../data/txts (YOLO)/3qf88IGrJRd8pNl1ShzqQQ.txt
19046 ../data/images/0012522.jpg ../data/txts (YOLO)/0012522.txt
19047 ../data/images/0002731.jpg ../data/txts (YOLO)/0002731.txt

19048 rows × 2 columns

In [70]:
import json
f = open('../data/classes.json')
data = json.load(f)
f.close()

labels = {}
for label in data:
    labels[data[label]['classIndex']] = label
In [73]:
!pip3 install pybboxes
import pybboxes as pbx

Requirement already satisfied: pybboxes in /Library/Frameworks/Python.framework/Versions/3.11/lib/python3.11/site-packages (0.2.0)
Requirement already satisfied: numpy==1.24.2 in /Library/Frameworks/Python.framework/Versions/3.11/lib/python3.11/site-packages (from pybboxes) (1.24.2)
Requirement already satisfied: pycocotools==2.0.6 in /Library/Frameworks/Python.framework/Versions/3.11/lib/python3.11/site-packages (from pybboxes) (2.0.6)
Requirement already satisfied: pyyaml==6.0 in /Library/Frameworks/Python.framework/Versions/3.11/lib/python3.11/site-packages (from pybboxes) (6.0)
Requirement already satisfied: matplotlib>=2.1.0 in /Library/Frameworks/Python.framework/Versions/3.11/lib/python3.11/site-packages (from pycocotools==2.0.6->pybboxes) (3.10.3)
Requirement already satisfied: contourpy>=1.0.1 in /Library/Frameworks/Python.framework/Versions/3.11/lib/python3.11/site-packages (from matplotlib>=2.1.0->pycocotools==2.0.6->pybboxes) (1.3.2)
Requirement already satisfied: cycler>=0.10 in /Library/Frameworks/Python.framework/Versions/3.11/lib/python3.11/site-packages (from matplotlib>=2.1.0->pycocotools==2.0.6->pybboxes) (0.12.1)
Requirement already satisfied: fonttools>=4.22.0 in /Library/Frameworks/Python.framework/Versions/3.11/lib/python3.11/site-packages (from matplotlib>=2.1.0->pycocotools==2.0.6->pybboxes) (4.58.2)
Requirement already satisfied: kiwisolver>=1.3.1 in /Library/Frameworks/Python.framework/Versions/3.11/lib/python3.11/site-packages (from matplotlib>=2.1.0->pycocotools==2.0.6->pybboxes) (1.4.8)
Requirement already satisfied: packaging>=20.0 in /Library/Frameworks/Python.framework/Versions/3.11/lib/python3.11/site-packages (from matplotlib>=2.1.0->pycocotools==2.0.6->pybboxes) (25.0)
Requirement already satisfied: pillow>=8 in /Library/Frameworks/Python.framework/Versions/3.11/lib/python3.11/site-packages (from matplotlib>=2.1.0->pycocotools==2.0.6->pybboxes) (11.2.1)
Requirement already satisfied: pyparsing>=2.3.1 in /Library/Frameworks/Python.framework/Versions/3.11/lib/python3.11/site-packages (from matplotlib>=2.1.0->pycocotools==2.0.6->pybboxes) (3.2.3)
Requirement already satisfied: python-dateutil>=2.7 in /Library/Frameworks/Python.framework/Versions/3.11/lib/python3.11/site-packages (from matplotlib>=2.1.0->pycocotools==2.0.6->pybboxes) (2.9.0.post0)
Requirement already satisfied: six>=1.5 in /Library/Frameworks/Python.framework/Versions/3.11/lib/python3.11/site-packages (from python-dateutil>=2.7->matplotlib>=2.1.0->pycocotools==2.0.6->pybboxes) (1.17.0)

[notice] A new release of pip is available: 24.0 -> 25.1.1
[notice] To update, run: pip3 install --upgrade pip
In [75]:
df2 = df.sample(20) # takes samples with bbox
y = 5; x = 2
plt.figure(figsize=(12.8*x, 7.2*y))
for idx in range(x*y):
    row = df2.iloc[idx]
    img = cv2.imread(row.image)
    dh, dw, _ = img.shape
    fl = open(row.label, 'r')
    data = fl.readlines()
    fl.close()

    for dt in data:
        cl, xc, yc, w, h = map(float, dt.split(' '))
        box_voc = pbx.convert_bbox((xc,yc,w,h), from_type="yolo", to_type="voc", image_size=(dw, dh))
        cv2.rectangle(img, (box_voc[0], box_voc[1]), (box_voc[2], box_voc[3]), (255, 0, 0), 5)
        img = cv2.putText(
          img = img,
          text = labels[cl],
          org = (box_voc[0], box_voc[1]),
          fontFace = cv2.FONT_HERSHEY_DUPLEX,
          fontScale = 2.0,
          color = (125, 246, 55),
          thickness = 3
        )
    plt.subplot(y, x, idx+1)
    plt.imshow(img)
    plt.axis('OFF')
plt.tight_layout()
plt.show()
No description has been provided for this image

Split dataset¶

In [ ]: